Synthesis and Applications of Small Molecule Libraries - Chemical

His laboratory is engaged in the development of new chemistry for the synthesis of organic compound libraries, and the application of organic compound...
1 downloads 0 Views 2MB Size
Chem. Rev. 1996, 96, 555−600

555

Synthesis and Applications of Small Molecule Libraries Lorin A. Thompson and Jonathan A. Ellman* Department of Chemistry, University of California, Berkeley, California 94720 Received August 17, 1995 (Revised Manuscript Received October 18, 1995)

Contents I. Introduction II. Libraries Synthesized on a Solid Support A. Library Synthesis and Evaluation Strategies 1. Discrete Compounds 2. Split Synthesis 3. Deconvolution of Soluble Libraries 4. Structural Determination by Analytical Methods 5. Encoding Strategies B. Synthesis of Organic Compound Libraries 1. Introduction to Solid Supports 2. Post-Synthesis Peptide Modification 3. Biopolymer-Mimetic Libraries 4. Nonoligomeric Compound Libraries 5. Molecular Recognition in Designed Receptor Systems 6. Analytical Techniques III. Libraries Synthesized in Solution A. Spatially Separate Synthesis B. Synthesis in Pools 1. A Library of Amides and Esters 2. Acetylcholinesterase Inhibitors 3. Amides Displayed from a Core Molecule 4. Oligosaccharide Libraries IV. Future Directions V. Acknowledgments VI. Glossary VII. Bibliography

555 556 556 556 557 558 559 560 561 561 562 562 566 590 591 592 592 593 593 594 595 595 596 597 597 597

Lorin A. Thompson was born in Lexington, KY, in 1970. He received the Bachelor of Science degree from the University of North Carolina, Chapel Hill, in 1992 where he worked under the guidance of Joseph Desimone. He is currently pursuing his doctorate in the laboratory of Jonathan Ellman at UC Berkeley where he is the 1994 Glaxo-Wellcome fellow. His research interests include the development of synthetic methodology for organic library construction.

I. Introduction One of the initial steps in the development of therapeutic agents is the identification of lead compounds that bind to the receptor or enzyme target of interest. Many analogs of these lead compounds are then synthesized to define the key recognition elements for maximal activity. In general, many compounds must be evaluated in both the lead identification and optimization steps. Increasing burdens have been placed on these efforts due to the large number of new therapeutic targets that continue to be identified thorough modern molecular biology methods.1 To address this demand, very powerful chemical and biological methods have been developed for the generation of large combinatorial libraries of peptides2 and oligonucleotides3 that are then screened against a receptor or enzyme to identify high-affinity ligands or potent inhibitors, respectively. While these studies have clearly demonstrated the power of library synthesis and screening strategies, peptides 0009-2665/96/0796-0555$25.00/0

Jonathan Ellman was born in California in 1962. He received his S.B. degree from M.I.T. in 1984 where he worked in the laboratory of K. Barry Sharpless. He received his Ph.D. degree with David A. Evans at Harvard University in 1989, where he worked on the synthesis of enantiomerically pure nonproteinogenic amino acids, cyclopeptide alkaloids, and vancomycin. After an NSF postdoctoral fellowship with Peter G. Schultz at the University of California at Berkeley on the incorporation of unnatural amino acids into proteins, he joined the faculty at the University of California at Berkeley in 1992 as an assistant professor. His laboratory is engaged in the development of new chemistry for the synthesis of organic compound libraries, and the application of organic compound libraries to different research problems in chemistry and biology.

and oligonucleotides generally have poor oral activities and rapid in vivo clearance;4 therefore their utility as bioavailable therapeutic agents is often limited. Due to the favorable pharmacokinetic properties of many small organic molecules (90% yield and purity. Permethylation of peptides that incorporate many of the different side chain functionalized amino acids was also investigated. As shown in Table 1, methylation of the functionalized side chains generally was observed, as were lower levels of purity of the permethylated products. For each of the peptides, as much as 15% of the terminal amino group had not been fully quaternized in the permethylation product. Longer reaction times could be employed to drive the quaternization to completion, but also resulted in appreciable degradation of the product. Houghten demonstrated that minimal epimerization occurs in the deprotonation step by treatment of all four possible stereoisomers of GGFL-NH2 with NaH in DMSO followed by a water quench. Subsequent cleavage from the support and HPLC analysis showed that less than 0.75% epimerization had occurred. Employing the above post-synthesis methylation strategy, Houghten performed the synthesis and evaluation of a library that theoretically contained 37 791 360 unique permethylated hexapeptides using the positional scanning approach (see section II.A.3). In particular, the library was screened for the ability to inhibit the growth of five different strains of bacteria or yeast. A number of permethylated peptides that incorporated multiple phenylalanines were identified that had significant activity against Staphylococcus aureus, methicillin-resistant S. aureus, and Staphylococcus sanguis, with IC50 values in the 1-10 µM range. A number of researchers have also reported on the post-synthesis modification of individual residues of peptides within a library, such as acylation or reductive amination of the N-terminus, or functionalization of reactive side chain functionality. However, these studies are outside the scope of this review and will not be discussed here.

3. Biopolymer-Mimetic Libraries Although peptides generally have poor pharmacokinetic properties that limit their utility as drugs, the high level of success that has been achieved in identifying high-affinity ligands to diverse receptors and enzymes through the synthesis and evaluation of peptide libraries is well documented.2 Many researchers have therefore focused on the synthesis and evaluation of biopolymer mimetics that although based upon the peptide structure, incorporate backbones that may have improved pharmacokinetic

Synthesis and Applications of Small Molecule Libraries

Chemical Reviews, 1996, Vol. 96, No. 1 563 Table 2. Pentamer Peptoids Prepared by the Submonomer Method (Scheme 2, Method B) pentamer side chain (R) (CH2)CH3 (CH2)Ph (CH2)Chx (CH2)CH(Ph)2 Ph c-C3H5 CH2)2indole (CH2)3NH2

characterization of unpurified product purity,a %

mass recovery,b %

>85 >85 >85 >85 >85 >85 >60 >85

90 74 79 70 83 83 52 63c

a Purity of unpurified compounds as determined by RPHPLC. b Determined from dry weight. c Prepared with BocNH-(CH2)3NH2.

Figure 9. Unnatural biopolymers.

properties as a result of proteolytic stability, more favorable solubility characteristics, or unique structural and/or hydrogen bonding motifs. Biopolymer mimetics are also of considerable interest to other research areas, including the design of two- and three-dimensional unnatural biopolymer frameworks with novel properties, including the design of molecular receptors (vide infra). Peptoids. Simon and co-workers considered a number of criteria for design of a new scaffold including simple synthesis of monomers, increased resistance to hydrolytic enzymes, the ability to display a wide range of functionality, high-yielding coupling steps amenable to automation, and the use of achiral monomers.49 Oligo(N-substituted)glycines, or “peptoids” (Figure 9) were proposed to meet these requirements. The side chains of peptoids are displayed from the amide nitrogen of an oligoglycine backbone instead of the R-carbon atom, providing a protease-resistant50 and achiral tertiary amine linkage. In the original synthesis of peptoids, standard solid-phase peptide synthesis methods were employed with Fmoc-protected N-alkylglycines serving as the monomer components (method A, Scheme 2).49 Scheme 2

(Benzotriazol-1-yloxy)tris(pyrrolidino)phosphonium hexafluorophosphate (PyBOP) or bromotris(pyrrolidino)phosphonium hexafluorophosphate (PyBroP) were found to be the optimal coupling agents, allowing the synthesis of up to 25-mers in high yield and excellent purity as judged by MS and RP-HPLC analysis. Subsequently, a simplified approach was reported by

Zuckermann and co-workers. In the submonomer approach (method B, Scheme 2),51 each cycle of synthesis involves a two-step procedure: (1) amide bond formation with R-bromo acetic acid employing 1,3-diisopropylcarbodiimide (DICI) as the activating agent and (2) bromide displacement with a suitable primary amine to provide the secondary N-alkylglycine (2) ready for the next coupling step. This approach allows for the direct incorporation of commercially available amines as building blocks; thereby eliminating costly, time-consuming monomer synthesis and obviating the need for R-amine protection. As shown in Table 2, the mass balance and purity of unpurified material is high for the synthesis of pentamers that incorporate a range of amine nucleophiles, including R-branched amines and unreactive aniline derivatives. In addition, a nonomer with five consecutive propylamine side chains followed by four consecutive butyl side chains (not listed in table) was obtained in greater than 65% purity as determined by RP-HPLC and in 86% crude mass balance, demonstrating the efficiency of the synthesis sequence. Using the submonomer approach, Zuckermann and co-workers assembled a library on Merrifield beads using the split-synthesis approach that was biased toward the 7-transmembrane G-protein coupled class of receptors.52 The Chiron group synthesized the library of approximately 5000 dimer and trimer peptoids employing 23 different monomers as well as three terminal amine capping reagents (Figure 10). Seven monomers were selected based upon the structures of known ligands to this class of receptors, while 17 monomers were selected to be as diverse as possible. The Chiron group has recently described their experimental design to maximize molecular diversity for a given library or to target a library with key features for a given receptor.53 Initial studies were performed to demonstrate that all of the amines were fully incorporated into a test oligomer. The library was then synthesized using a Zymark robot to carry out resin manipulations and micropipetting54 and assayed as 18 pools of 216, 255, or 272 compounds for the ability to inhibit [3H]prosazin binding to an R1-adrenergic receptor preparation. By using the iterative resynthesis and evaluation strategy, a 5 nM inhibitor was identified (3, Figure 11). Similarly, a 6 nM inhibitor of [3H]DAMGO binding to a µ-specific opioid receptor preparation was identified (4, Figure 11). The peptoid products are completely stable to proteolysis as

564 Chemical Reviews, 1996, Vol. 96, No. 1

Thompson and Ellman

pared by reduction of the corresponding protected amino acids followed by treatment of the resulting alcohols 5 with p-nitrophenyl chloroformate (Scheme 3). Oligomers were initially assembled on polystyScheme 3

Figure 10. Monomer side chains and capping reagents used in peptoid library synthesis.

Figure 11. Peptoid ligands.

determined by incubating a number of peptoid derivatives with carboxypeptidase A, chymotrypsin, elastase, papain, pepsin and thermolysin.50b Additional diversity has also been achieved through modification of peptoid side chains. Pei and Moos have described a procedure for modification of peptoids that have alkenyl or alkynyl side chains by a [3 + 2] cycloaddition reaction with nitrile oxides to produce support-bound isoxazoles and isoxazolines.55 The products are formed with high regioselectivity in >80% purity by HPLC analysis, and the authors conclude by the lack of byproducts that the chemical yields are similar, although only one mass balance (72%) of an unpurified product is provided. The Chiron group has also reported multistep modification of peptoid structures to provide six- and sevenmembered ring heterocycles. These studies are described in later sections. Oligocarbamates. Another unnatural biopolymer-based library was reported by Cho and coworkers.17,56 Their method involves the use of amino acids as precursors to chiral building blocks for the synthesis of a library of oligocarbamates. N-Protected p-nitrophenyl carbonate monomers 6 were pre-

rene resin using the Fmoc group as the amine protecting group and a standard Rink amide linker. Two reactions were performed in each coupling cycle: (1) removal of the Fmoc group by treatment with 20% piperidine in N-methylpyrrolidinone (NMP) and (2) coupling with an N-Fmoc carbonate monomer employing hydroxybenzotriazole (HOBt) as an additive. Coupling yields were determined to be >99% by RP-HPLC and quantitative ninhydrin tests,57 and oligocarbamates were characterized by FAB-MS and 1H NMR spectroscopy. Two representative oligocarbamates were also incubated with trypsin and porcine pepsin and were found to be stable to proteolysis. The photolithography technique developed by Fodor and co-workers (section II.A.1.b) was then used to prepare a library of oligocarbamates.17 In order to introduce a protecting group scheme that was compatible with photolithography, the carbamate monomers were prepared with the photolabile [(nitroveratryl)oxyl]carbonyl (Nvoc) group in place of the Fmoc group. An eight-step binary masking scheme was then employed to synthesize a library of 256 oligocarbamates around the parent sequence AcYcFcAcScKcIcFcLc (where Xc refers to the carbamate monomer formed from the amino acid X) such that the library contained all of the possible deletion sequences for the parent oligomer. The library was assayed for the ability to bind the monoclonal antibody (mAb) 20D6.3, which was raised to the keyhole limpet hemocyanin conjugate of oligocarbamate AcYcKcFcLcG-OH (G-OH is a terminal glycine residue). Binders were identified by scanning epifluorescent microscopy using a goat R-mouse fluorescein-conjugated secondary antibody. Five out of the 10 highest affinity oligocarbamates, AcKcFcLcG-OH, AcFcKcFcLcG-OH, AcYcKcFcLcG-OH, AcAcKcFcLcG-OH, and AcIcFcLcG-OH, were resynthesized and purified on large scale using Fmoc chemistry. The IC50 values of all five ligands determined in solution were in the 60-180 nM range. These studies also revealed that the support or linker can interfere with receptor interactions, since the ligand AcYcFcLcG-OH was also

Synthesis and Applications of Small Molecule Libraries

prepared on large scale and assayed in solution and found to have an IC50 value of approximately 160 nM, even though this ligand ranked in the bottom 30% of ligands in the support-bound assay. Oligoureas. Burgess has reported on initial studies toward the solid-phase synthesis of oligoureas with the goal of synthesizing oligourea libraries (Scheme 4).58 The monomers 10 for oligourea syn-

Chemical Reviews, 1996, Vol. 96, No. 1 565 Scheme 5

Scheme 4

thesis were prepared on average in 50-60% overall yield by the three step process of reduction of the corresponding N-Boc-protected amino acid followed by converting the resulting primary alcohol 9 to the phthalimide under Mitsunobu reaction conditions and final removal of the Boc protecting group. In the solid-phase synthesis of the oligourea, two reactions are performed in each coupling cycle: (1) removal of the phthaloyl group by treatment with 60% hydrazine hydrate in DMF and (2) coupling with a monomer that is activated in situ as the isocyanate by treatment with triphosgene. Two oligourea/peptide hybrids, 13 and 14 (Figure 12), were prepared by this general approach in unoptimized 46% and 17% isolated yield, respectively.

protected amino aldehyde (Scheme 5). A WittigHorner-Emmons reaction provides the R,β-unsaturated sulfonate 15, which is then converted into the sulfonyl chloride by deprotection with tetrabutylammonium iodide followed by activation with SO2Cl2 and PPh3.60 In the solid-phase synthesis of the sulfonyl peptides, two reactions are performed in each coupling sequence: (1) removal of the Boc group by treatment with trifluoroacetic acid (TFA) and (2) four cycles of coupling with 1 equiv of vinylsulfonyl chloride 16 followed by slow addition of 1,8-diazabicyclo[5.4.0]undecane (DBU). Four cycles of coupling and slow addition of DBU is necessary, since excess DBU results in decomposition of the vinylsulfonyl chloride monomers, but DBU is the only base that was studied that provides clean conversion to product. By using Tentagel resin, all five monomers, 16a to 16e, were coupled to the support-bound Gly ester followed by cleavage from the support by treatment with 10% triethylamine in methanol. The methyl ester products were isolated in good overall yield, 60-70%. In addition, two sulfonyl dipeptides were prepared to demonstrate oligomer synthesis by the above method. Sulfonyl dipeptides 19 and 20 were obtained in 57% and 52% isolated yields, respectively (Figure 13). In addition, only one diastereomer of 19 was detected indicating that no racemization occurred either in monomer synthesis or in the coupling steps.

Figure 12. Oligurea/peptide hybrids prepared on solid support.

Vinylogous Sulfonyl Peptides. Gennari and Still have reported on the solid-phase synthesis of vinylogous sulfonyl peptides with the goal of making sulfonyl peptide libraries as well as in employing these biopolymer mimetics as synthetic receptors.59 The authors focused upon the vinylogous sulfonamide structure because the sulfonamide may mimic the tetrahedral geometry of amide hydrolysis in the protease cleavage of peptide bonds thereby serving as an interesting pharmacophore for protease inhibition. In addition, the stronger polarization of the sulfonamide bond compared to a regular peptide bond favors the formation of hydrogen bonds to potentially provide a more distinct preorganization of the sulfonyl peptides. The monomers 16 for oligomer synthesis were prepared on average in 65-75% overall yield by a three step process from the corresponding N-Boc-

Figure 13. Vinylogous sulfonyl peptides.

Vinylogous Peptides. In one of the earliest studies on unnatural biopolymers, Schreiber and coworkers reported on the synthesis of vinylogous peptides and demonstrated that these structures can adopt specific secondary as well as tertiary structures (Figure 14).61 The vinylogous peptide 21 was observed to adopt a stacked array of parallel sheets as determined by X-ray crystallography, while the pep-

566 Chemical Reviews, 1996, Vol. 96, No. 1

Figure 14. Vinylogous peptide and vinylogous peptide/ amide hybrids.

tide/vinylogous amide hybrids 22 and 23 were determined to adopt structures corresponding to antiparallel sheets and helical conformations, respectively, with turns about the Pro-Gly bonds as determined by NMR. In this study, the compounds were prepared in solution for the purpose of structural evaluation; however, the monomer synthesis is expedient and solid-phase synthesis strategies for the purpose of library construction should be straightforward using standard peptide synthesis methods. Smith and Hirschmann have developed an unnatural biopolymer based upon linked pyrrolin-4-ones that also incorporates a vinylogous amide into its structure (Figure 9).62 Although these oligomers have not been employed for library synthesis, they do adopt a secondary structure that mimics a β-strand62c and have been utilized in the synthesis of potent protease inhibitors.62b

4. Nonoligomeric Compound Libraries Nonoligomeric molecules, which are nonpeptidic in nature and are below 600-700 in molecular weight, have become the major focus of library synthesis efforts for the development of medicinal agents. Libraries of small nonoligomeric molecules have been prepared both for the identification of lead compounds and for the optimization of lead compounds that have been identified through either library screening efforts or alternative methods. While the compound class for library synthesis toward lead optimization is predetermined, a number of factors must be considered in the selection of a compound class for the purpose of lead identification. To varying degrees, three general strategies for compound selection have been used for the library syntheses outlined in this review.63 The first strategy is to select “privileged” structures,64 where the display of different functionality upon the structure has previously provided a number of potent and specific therapeutic agents or candidates toward different therapeutic targets. The second strategy is to design compound scaffolds based on important recognition elements of biological receptors. The final strategy is to select stable compounds upon which few therapeutic agents or candidates have been based, but which are straightforward to prepare with multiple sites available for the display of functionality.

Thompson and Ellman

In designing a synthesis scheme to access a library based upon a specific compound class three criteria should be considered: (1) The synthesis scheme should provide the majority, if not all, of products in the library in high yield and purity. (2) The chemistry should be compatible with the display of as much diverse functionality as possible including heteroatom functionality that is commonly found in drugs such as alcohols, phenols, amines, indoles, guanidines, carboxylic acids, amides, nitriles, imidazoles, nitro groups, and halides. (3) The building blocks for synthesis of the library should be commercially available or at least readily accessible, since a library cannot be made rapidly and efficiently if many of the building blocks must be prepared.65 Finally, it should be emphasized that for a given compound class, we and others have found that developing a high-yielding and general synthesis sequence followed by rigorously establishing that a diverse array of functionality can be displayed can require significant effort. In contrast, using the optimized synthesis sequence libraries can be constructed rapidly and efficiently as long as the building blocks are commercially available or readily accessible. a. Heterocycle Libraries. 1. Seven-membered Rings. All of the efforts toward the synthesis of libraries of seven-membered ring heterocycles have focused upon the 1,4-benzodiazepine structure. The 1,4-benzodiazepine class of compounds have widespread biological activities and are one of the most important classes of bioavailable therapeutic agents. In addition to 1,4-benzodiazepines such as Valium that have anxiolytic activity,66 there are also derivatives that are highly selective cholecystokinin (CCK) receptor subtype A antagonists, highly selective CCK receptor subtype B antagonists, 67 κ-selective opioids,68 platelet-activation factor antagonists,69 HIV Tat antagonists,70 reverse transcriptase inhibitors,71a gpIIbIIIa inhibitors,71b and ras farnesyltransferase inhibitors.72 1,4-Benzodiazepin-2-ones. In one of the first articles to address the synthesis and evaluation of small molecule combinatorial libraries, Bunin and Ellman reported the solid-phase synthesis of 1,4benzodiazepine derivatives.73 In this initial report, benzodiazepine derivatives were constructed from three components: 2-aminobenzophenones, amino acids, and alkylating agents. By employing solution chemistry, a hydroxyl-substituted 2-N-Fmoc-aminobenzophenone is coupled to the [4-(hydroxymethyl)phenoxy]acetic acid (HMP) linker.74 The linkerderivatized aminobenzophenone 24 (Scheme 6) is Scheme 6

then coupled to the solid support by employing standard amide bond-forming methods (the linker may be attached to either ring of the 2-aminobenzophenone). Synthesis of the benzodiazepine derivative on solid support then proceeds by removal of the Fmoc

Synthesis and Applications of Small Molecule Libraries

Chemical Reviews, 1996, Vol. 96, No. 1 567

Scheme 7

Table 3. 1,4-Benzodiazepine Derivatives 30 (Scheme 7) derivative entry

R1

R2

R3

1 2 3 4 5 6 7 8 9 10

4′-OH 4′-OH 4′-OH 4′-OH 4′-OH 4′-OH 4′-OH 4′-OH

5-Cl 5-Cl 5-Cl 5-Cl 5-Cl 5-Cl 5-Cl 5-Cl 4-CO2H,5-Cl 4-CO2H,5-Cl

CH3 CH3 CH3 CH3 CH(CH3)2 CH2CO2H (CH2)4NH2 CH2Ph(4-OH) CH2Ph CH3

R4

yield (%)a

H 95 CH3 100 CH2CH3 97 CH2CHdCH2 90 CH2CH3 85 CH2CH3 95 CH2CH3 95 CH2CH3 98 CH3 100 CH2Ph 93

a Yields of purified material based on support-bound starting material 26 (Scheme 7).

protecting group from 26 by treatment with piperidine in DMF followed by coupling an R-N-Fmoc amino acid fluoride to the resulting unprotected 2-aminobenzophenone (Scheme 7). The activated R-N-Fmoc amino acid fluoride75 is employed in order to achieve complete conversion to the amide product 27 even for electron-deficient 2-aminobenzophenone derivatives. The Fmoc protecting group is then removed, and the resulting free amine is treated with 5% acetic acid in NMP to provide the initial benzodiazepine derivative 28. Alkylation of the anilide of 28 then provides the fully derivatized 1,4-benzodiazepine 29. To maximize synthesis generality, lithiated 5-(phenylmethyl)-2oxazolidinone76 or lithiated acetanilide is employed as the base since it is basic enough to completely deprotonate the anilide of 28, but not basic enough to deprotonate amide, carbamate, or ester functionality. By employing these conditions 1,4-benzodiazepine derivatives containing esters and carbamates were alkylated in high yield on solid support with no overalkylation observed (entries 6 and 7, Table 3). Treatment with standard trifluoroacetic acid cleavage reagents affords the benzodiazepine products 30 in high yield (85-100%, 95% av after purification based on support-bound starting material 26). Finally, no racemization of the amino acid component is detected (85% purity by 1H NMR analysis of crude products. Yields of purified benzodiazepine products varied from 46% to 72% (av 61%, 15 compounds) based on the initial aminomethyl loading of the polystyrene resin used.82 Bunin and co-workers have reported the preparation of a library of 11 200 discrete 1,4-benzodiazepines11 from 20 acid chlorides, 35 amino acids, and 16 alkylating agents, all of which were commercially available (Figure 16). The acid chlorides were selected from a set of over 300 commercially available acid chlorides that are compatible with the synthesis sequence using a similarity grouping procedure developed by Dr. Steven Muskal at MDL Information Systems to select as diverse a set as possible. The alkylating agents and amino acids also displayed a range of functionality. The library is currently being screened by a number of industrial and academic collaborators. Plunkett and Ellman have also demonstrated a silyl linkage strategy for the synthesis of benzodiazepine derivatives.83 Linkage of aromatic compounds to solid supports through silyl groups is likely to become an important and general strategy, since cleavage from the resin is accomplished by protodesilation to leave no trace of the linkage site. For the synthesis of 1,4-benzodiazepine derivatives, the silylsubstituted (aminoaryl)stannane derivative 34, which is synthesized in five steps in solution, is coupled to aminomethylated polystyrene (Scheme 9). The synScheme 9

thesis of support-bound 1,4-benzodiazepines then proceeds as described previously. Treatment of the

Synthesis and Applications of Small Molecule Libraries

Chemical Reviews, 1996, Vol. 96, No. 1 569 Scheme 10

cleavage (Scheme 10). The resins were purchased as Boc-protected Merrifield resins and deprotected with 1:1 TFA/CH2Cl2. The 2-aminobenzophenone imine derivatives 38 were then added to form the supportbound imines, followed by heating to 60 °C in TFA to provide the desired benzodiazepine product. The cyclative nature of the cleavage was designed to enhance the purity of the cyclized product, since incomplete products would be expected to stay bound Table 5. 1,4-Benzodiazepin-2-one Derivatives 40 (Scheme 10)

Figure 16. Building blocks used to prepare a 11 200member benzodiazepine library.

support-bound benzodiazepine 36 with anhydrous HF then provides the benzodiazepine product 37. Good purity of the crude product is observed, >85% by 1H NMR, and the isolated yield is 50-68% (av 61%, four compounds) after purification based on the initial aminomethyl loading of the polystyrene resin. In one of the early reports of small molecule library synthesis, DeWitt and co-workers described an alternative strategy for the synthesis of 1,4-benzodiazepin-2-ones.12 The synthesis is based on a two-step procedure of trans-imidation followed by a cyclative

entry

R1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

CH3 CH3 CH3 CH3 CH3 CH3 CH3 CH3 H H H H H H H H CH2C6H5 CH2C6H5 CH2C6H5 CH2C6H5 CH2C6H5 CH2C6H5 CH2C6H5 CH2C6H5 3-CH2indolyl 3-CH2indolyl 3-CH2indolyl 3-CH2indolyl 3-CH2indolyl 3-CH2indolyl 3-CH2indolyl 3-CH2indolyl CH(CH3)2 CH(CH3)2 CH(CH3)2 CH(CH3)2 CH(CH3)2 CH(CH3)2 CH(CH3)2 CH(CH3)2

derivative R2 C6H5 C6H5 C6H5-4-OCH3 C6H5 see b below C6H5 c-C6H11 2-thienyl C6H5 C6H5 C6H5-4-OCH3 C6H5 see b below C6H5 c-C6H11 2-thienyl C6H5 C6H5 C6H5-4-OCH3 C6H5 see b below C6H5 c-C6H11 2-thienyl C6H5 C6H5 C6H5-4-OCH3 C6H5 see b below C6H5 c-C6H11 2-thienyl C6H5 C6H5 C6H5-4-OCH3 C6H5 see b below C6H5 c-C6H11 2-thienyl

R3

R4

mass balance (%)a

H Cl H NO2 see b below Cl H H H Cl H NO2 see b below Cl H H H Cl H NO2 see b below Cl H H H Cl H NO2 see b below Cl H H H Cl H NO2 see b below Cl H H

H H H H H Me H H H H H H H Me H H H H H H H Me H H H H H H H Me H H H H H H H Me H H

40 56 34 28 63 18 41 47 44 55 23 31 16 20 32 41 52 46 41 26 52 13 39 48 43 33 31 23 23 10 34 40 31 28 29 9 29 11 27 37

a Mass balance of crude material cleaved from the resin based on amino acid loading level. b Prepared from 2-amino9-fluorenone rather than a 2-aminobenzophenone.

570 Chemical Reviews, 1996, Vol. 96, No. 1

Thompson and Ellman Table 6. Synthesis of 1,4-Benzodiazepine-2,5-diones 46 (Scheme 11)

Figure 17. Structures of 1,4-benzodiazepin-2-one and 1,4benzodiapine-2,5-dione.

to the support. The desired benzodiazepine derivatives were purified by extraction and isolated in crude mass balances of 9-63% (av 34%, 40 compounds) in >90% purity as measured by 1H NMR (Table 5). 1,4-Benzodiazepine-2,5-diones. Boojamra and co-workers have also reported a general and highyielding method for the solid-phase synthesis of 1,4benzodiazepine-2,5-diones (Figure 17).84 The synthesis strategy complements the previously described 1,4-benzodiazepin-2-one synthesis sequence described by Bunin and co-workers, since a wide range of functionality can be directly introduced onto the aromatic core of the benzodiazepine structure (R1) from the greater than 40 commercially available anthranilic acids or related heterocyclic structures. The other two sites of diversity are introduced with R-amino esters and alkylating agents, of which there are also a number of derivatives that are commercially available. The synthesis of 1,4-benzodiazepine-2,5-diones is initiated by loading an R-amino ester onto the aldehyde-derivatized support by reductive amination employing NaBH(OAc)3 in DMF with 1% AcOH (Scheme 11).85 Racemization is not observed if the Scheme 11

entry

R1

R2

R3

yield (%)a

1 2 3 4 5 6 7b 8b 9 10 11

8-Cl 7-Cl 7-Br 8-NO2 6-F 8-OCH3 7-(C6H4)-p-OCH3 8-(CH2)5CH3 7-Cl 8-Cl 7-Cl

CH2CH3 CH2CHdCH2 CH2CH3 CH2CH3 CH2CONH2 CH2c-C3H5 CH2CH3 CH2C6H4-p-Ph CH2CHdCH2 H H

CH2CH(CH3)2 CH2C6H5 CH2CH(CH3)2 CH2CH(CH3)2 CH2C6H5 CH2CH(CH3)2 CH2CH(CH3)2 CH2C6H4OH (CH2)4NH2 CH2CH(CH3)2 CH2C6H5

75 89 71 92 62 79 62 77 63 89 89

a Yields of purified materials are based on the loading levels of leucine and phenylalanine ester derived resins. b Suzuki cross-coupling products.

carbodiimide (EDC) is the most convenient activating agent since the tertiary amine hydrochloride is present in the carbodiimide structure. Cyclization and then subsequent alkylation of the support-bound anilide anion 44 generated in situ is next accomplished in a single step by treatment of amide 43 with the lithium salt of acetanilide in DMF/THF (1:1) for 30 h, followed by addition of an appropriate alkylating agent. Additional diversity may also be introduced onto the benzodiazepine through the Suzuki cross-coupling reaction as is exemplified in entry 7 (Table 6), where a cross-coupling reaction was carried out with p-methoxybenzeneboronic acid, and in entry 8 (Table 6) where a Suzuki cross-coupling reaction was performed using B-hexyl-9-BBN. The benzodiazepine products are cleaved from the support by treatment with TFA/Me2S/H2O (90:5:5). Good yields were obtained for a range of different derivatives including benzodiazepines that incorporate amino acids with side-chain functionality such as tyrosine and lysine, entries 8 and 9 in Table 6, respectively. In addition, no detectable racemization, 90 41 53 34 68 41 52 60 50 90 37 nd nd 50 75 55 41 58 49

>65 80 79 92 >95 80 61 72 69 70 97 63 61 88 93 59 93 84 85 83 64

a Yields of unpurified product based on the loading level of the starting resin. b Purity as estimated by HPLC. c Derived from myrtanylamine. d Derived from 5-aminoindan. nd ) not determined.

The resulting amine 48 is then acylated with a substituted o-azidobenzoyl chloride. Reaction with tributylphosphine affords the iminophosphorane, which cyclizes upon heating to 125 °C to afford the support-bound benzodiazepine. Cleavage with TFA provides the 1,4-benzodiazepine-2,5-dione 50, which is lyophilized twice from acetic acid. A number of amino acids were evaluated for compatibility with the synthesis sequence (entries 1-12, Table 7). The study includes the synthesis of 21 compounds that were isolated in crude mass balances of 34-90% (av 55%) and HPLC purities estimated at 59-97% (av 78%). Racemization was examined for one model benzodiazepine (entry 18). The diastereomeric excess of this product was 87%, and further studies are reported to be in progress. A model study was also performed to demonstrate that the synthesis sequence could be performed in a split synthesis format for library synthesis. An equimolar mixture of eight monopeptoid resins were pooled and transformed into 1,4-benzodiazepine-2,5diones using L-Phe methyl ester and 2-azidobenzoyl chloride. Cleavage of the resulting products from the resin provided a sample pool of eight compounds that was evaluated by Electrospray MS and HPLC. All eight desired benzodiazepines were present in addition to uncyclized material. The authors report that further functionalization of the benzodiazepin-2,5diones can be accomplished by Suzuki reaction of triflate substituted derivatives or a nitro reduction and acylation sequence (data not provided). 2. Six-membered Rings. Diketopiperazines. Gordon and Steele have reported a general strategy for the solid-phase synthesis of diketopiperazines employing readily available amino acids and aldehydes to introduce diversity.87 Reductive amination of a support-bound amino acid was first accomplished by treatment with an aldehyde and sodium triacetoxyborohydride in CH2Cl2 (Scheme 13). The reductive amination step was performed twice to improve conversions to 85-95%, except for sterically hindered

Scheme 13

amino acids and/or electronically deactivated aldehydes which gave lower conversions. In fact, for the most difficult case where Val and an electronically deactivated aldehyde was employed, only 20% conversion was observed. Partial racemization was also noted for some amino acids (Phe gave approximately 10% racemization by chiral HPLC). Both aromatic and aliphatic aldehydes were employed in the reductive amination step, although aliphatic aldehydes provided noticeable dialkylation (1-10%). An N-Boc amino acid was then coupled to the secondary amine employing PyBroP as the activating agent (double coupling was used to drive the reaction to completion). The Boc protecting group was then removed with concomitant cleavage from the support with TFA. The resulting dipeptide acids were then dissolved in toluene and heated at reflux for 5 h to provide the desired diketopiperazine products 54. The yields of two purified diketopiperazines were reported, 42% yield when R1 ) CH2C6H5, R2 ) CH2C6H44-OCH3, and R3 ) CH3, and 24% yield when R1 ) CH(CH3)2, R2 ) CH2C6H2-2,4,6-(OCH3)3 and R3 ) CH2CH(CH3)2. A library of 1000 DKP’s was then prepared using the split synthesis approach from 10 N-Fmoc amino acid-derivatized resins, 10 aldehydes, and 10 N-Boc amino acids (Figure 18). The 10 N-Fmoc amino acidderivatized resins were deprotected, pooled, and

572 Chemical Reviews, 1996, Vol. 96, No. 1

Thompson and Ellman Table 8. 2-Substituted 1-(2H)-Isoquinolinones 58 (Scheme 14) derivative

Figure 18. Building blocks used to prepare a 1000member diketopiperazine library.

entry

R1

R2

yield (%)a

purity (%)b

1 2 3 4 5 6 7 8

CH2CH(CH3)2 CH2CH2C6H5 C6H5 CH2CH(CH3)2 CH2CH(CH3)2 CH2CH(CH3)2 CH2CH(CH3)2 CH2CH(CH3)2

H H H 5-CH3 8-F 6,7-(OCH3)2 7-Cl 5-OCH3

69 65 85 92 (1/3.2)c 80 77 79 69 (1.7/1)c

83 80 >70 94 90 95 90 93

a Yields of unpurified product after lyophilization from acetic acid. b Purity as estimated by HPLC. c Values in parentheses are ratios of 58 to 59, otherwise only 58 was observed.

Scheme 14

reductively alkylated with 10 different aldehydes to generate 10 mixtures of 10 support-bound N-alkyl amino acids 52. These mixtures were characterized using HPLC-MS and MS-MS after cleavage from the support to confirm the presence of 96 out of the 100 expected N-alkyl amino acids. The resulting resins were then repooled, mixed, and divided into 10 separate pools. Each pool was coupled with a unique N-Boc amino acid, followed by cleavage from the resin and cyclization to provide a library of 1000 compounds as 10 mixtures of 100 compounds each. In a later publication, Terrett reported that biological evaluation of this library resulted in the identification of a number of active diketopiperazines including a ligand to the neurokinin-2 receptor (IC50 ) 313 nM).6b Isoquinolinones. Zuckermann and Goff have reported a method for the solid-phase synthesis of isoquinoline derivatives, which grew from a desire to increase the structural rigidity, complexity, and diversity of their existing peptoid libraries.88 Coupling of trans-4-bromo-2-butenoic acid to the support through the Rink linker is accomplished with DICI (Scheme 14). Bromide displacement with a primary amine is followed by acylation of the resulting secondary amine with a 2-iodobenzoyl chloride derivative to provide the support-bound amide 57. Palladium-mediated cyclization by an intramolecular Heck reaction using Pd(PPh3)4 followed by cleavage with TFA then provides the (2H)-isoquinoline 58. However, when a 2-iodobenzoyl chloride derivative was employed that also contained a substituent in the 3-position (entries 4 and 8, Table 8), a mixture of product isomers 58 and 59 was seen, which could favor isomer 59 (entry 4, Table 8). Further experiments lead the authors to conclude that isomer 59 is produced first, with equilibration to isoquinoline 58 under the reaction conditions, probably through a mechanism of readdition and elimination of Pd-H. Presumably, an ortho substituent hinders the readdition of Pd-H. Mass balances after cleavage and lyophilization are reported for eight compounds (6592%, av 77%), which are obtained in good to purity as estimated by HPLC (70-95%, av 87%). A pooling strategy was also investigated, where seven of the different support-bound intermediates 56 (Scheme 14) were mixed and then converted to the isoquinoline products. Cleavage and HPLC showed the presence of all seven desired products and the structures were verified by ES-MS.

1,4-Dihydropyridines. Gordeev and co-workers at Affymax have utilized a Hantzsch-type reaction to provide 1,4-dihydropyridines, which have served as the nucleus of numerous bioactive compounds.89 Condensation of a β-keto ester with the free amine from either the Rink (shown) or the PAL linker gives the support-bound enamino ester 60 (Scheme 15). Treatment with an aldehyde and a second β-keto ester, which forms a 2-arylidene β-keto ester in situ (this reagent can also be preformed), provides the resin-bound intermediate 61. Cleavage with TFA results in cyclization to give 1,4-dihydropyridines 62 (Table 9). The intermediacy of a support-bound cyclized 1,4-dihydropyridine is also possible; however, in a model study using 13C-enriched ethyl acetoacetate (R1 ) 13C methyl), the 13C NMR of the supportbound intermediate prior to cyclization showed two signals for the R1 methyl, tentatively assigned to the Scheme 15

Synthesis and Applications of Small Molecule Libraries

Chemical Reviews, 1996, Vol. 96, No. 1 573

Table 9. 1,4-Dihydropyridine Derivatives 62 (Scheme 15) derivative entry

Ar

R1

R2

R3

R4

yield (%)a

1 2 3 4 5 6 7 8 9

C6H4-2-NO2 C6H4-3-NO2 C6H4-3-NO2 C6H4-4-NO2 C6H4-4-NO2 C6H4-4-NO2 C6H4-4-CN C6H5 4-pyridyl

CH3 CH3 CH3 CH3 CH3 CH2CH3 CH3 CH3 CH3

CH3 CH3 CH(CH3)2 CH3 CH2CH3 CH3 CH3 CH3 CH3

CH3 CH3 CH3 CH3 CH3 C6H5 CH3 CH3 CH3

OCH3 OCH2CH3 O(CH2)2OCH3 CH3 OCH3 OCH2CH3 OCH3 OCH2CHdCH2 OCH3

65b (70)c 75c 78c 75c 70c 70c 74c 72b 75c

a Yields are based on the loading level of the starting resin. b Prepared by two-component condensation with ArCHdC(COR )CO R 3 2 4 using PAL resin. c Prepared by three-component condensation using Rink resin, ArCHO, and R3COCH2COR4.

E and Z isomers of intermediate 61. The IR of the support-bound intermediate also showed an absorbance at 1735 cm-1 while the cyclized product showed the expected absorbance at 1705 cm-1. Interestingly, pyridine was also necessary in the condensation step to facilitate isomerization of the imine to the enamine, as undesired cycloaddition byproducts were observed without it. The synthesis sequence is notable because the reaction conditions are mild, the synthesis proceeds in good overall yields (av 73%, nine compounds), and a variety of functionality could potentially be displayed about the structure from commercially available starting materials. Dihydro- and Tetrahydroisoquinolines. Meutermans and Alewood have demonstrated the synthesis of dihydro- and tetrahydroisoquinolines on support.90 Acylation of support-bound dimethoxyphenylalanine with either acetic acid or phenylacetic acid produced the intermediate amide 64 or 65 (Scheme 16). The support-bound dihydroisoquinolines 66 and 67 were then obtained by treatment of amides 64 and 65 with POCl3 at 80 °C to effect a Bischler-Napieralski cyclization reaction. The dihydroisoquinolines 68 and 69 were released from the resin by treatment with HF/p-cresol. The supportbound dihydroisoquinolines 66 and 67 were also converted to the tetrahydroquinolines 70 and 71 by treatment with NaBH3CN before cleavage from the support. All four products were obtained in high purity as estimated by HPLC analysis. The isolated

yield after HPLC purification for each of the dihydroisoquinolines 68 and 69 was ∼40%, and the yields of the tetrahydroisoquinolines 72 and 73 were 25% and 30%, respectively, as a 6:1 mixture of diastereomers for both compounds. In a separate experiment an equimolar mixture of eight acetic acid derivatives (R1 ) H, phenyl, t-Bu, naphthyl, p-methoxyphenyl, 3,4-dimethoxyphenyl, p-nitrophenyl, and p-hydroxyphenyl) was coupled to support-bound amino ester 63 and then submitted to the cyclization and cleavage steps. The parent ions of all eight dihydro- and tetrahydroisoquinolines were identified by ionspray MS, but no other characterization of the mixtures was provided. 3. Five-Membered Rings. Hydantoins. In one of the earliest studies of organic compound library synthesis, DeWitt and co-workers reported a method for the synthesis of hydantoin derivatives employing the DiversomerTM apparatus.12 Eight resin-bound amino acids (Phe, Gly, Ile, Leu, Ala, Val, Trp, and diphenylglycine) were condensed with five different isocyanates (structures not provided) in DMF to provide resin-bound ureas 74 (Scheme 17). Heating Scheme 17

Scheme 16

the resin-bound ureas 74 at 85-100 °C in 6 M HCl for 2 h resulted in cyclative cleavage to provide the hydantoin products 75 in crude mass balances of 4-81%, (av 30%, 40 compounds) with characterization by TLC, MS, and 1H NMR. Pyrrolidines. Gallop and co-workers have developed a solid-phase synthesis of mercaptoacyl proline derivatives with a metalloazomethine ylide cycloaddition reaction serving as the key step for introducing diversity.91 Condensation of an amino acid-derivatized resin with an aromatic aldehyde in neat trimethylorthoformate provided the Schiff base 76, and any unreacted amines were capped with acetic anhydride (Scheme 18). The dehydrating conditions that were employed to effect Schiff base formation

574 Chemical Reviews, 1996, Vol. 96, No. 1

Thompson and Ellman

Scheme 18

Figure 20. Captopril and the identified mercaptoacyl proline ACE inhibitor, 79.

had previously been reported by the authors.92 The 1,3-dipolar cycloaddition was then performed with the azomethine ylide derived from the support-bound Schiff base 76 and an acrylate or acrylonitrile as the dipolarophile under Lewis acid-mediated conditions that are analogous to those initially reported by Grigg and Tsuge for the corresponding solution-phase reaction.93 The resulting racemic substituted proline derivatives were then acylated with various ω-mercaptoacyl chlorides followed by acidolytic cleavage from the support to provide the desired (mercaptoacyl)proline product 78 (Scheme 18). The isolated yields for six representative products ranged from 50 to 80% and the diastereoselectivities ranged from 2.5:1 to 10:1. The authors reported that the products typically arose from an endo-selective cycloaddition with the syn configuration of the support-bound azomethine ylide. Dipolarophiles lacking a carbonyl substituent showed decreased stereoselectivity. In addition, the diastereoselectivity was dependent upon the nature of the resin support and the concentration of the Lewis acid catalyst. The authors prepared a small library of (mercaptoacyl)proline derivatives 78, employing the optimized synthesis sequence and the split synthesis procedure. The library was prepared from four amino acids, four aromatic aldehydes, five olefins, and three mercaptoacyl chlorides (Figure 19). Because the library was targeted toward the identification of angiotensin-converting enzyme (ACE) inhibitors, the building blocks were selected in part upon the structure activity relationships of known (mercaptoacyl)proline-based inhibitors such as captopril. Although the total number of combinations of the building blocks in the library was 240; because the reactions were not completely regio- and stereospecific, approximately 500 compounds were prepared.

The library was deacetylated by treatment with ethylenediamine and then screened for inhibition of ACE. Deconvolution using the standard iterative resynthesis and evaluation protocol resulted in the identification of a potent new inhibitor, 79 (Figure 20), with a Ki of 160 pM. Inhibitor 79 is 3-fold more potent than captopril and is among the most potent thiol-containing ACE inhibitors yet described. Thiazolidine-4-carboxylic Acids. Utilizing FmocCys(Trt)-OH as a starting point, the Selectide group has synthesized a number of N-acylthiazolidines on PEG-PS resin.94 Fmoc-Cys(Trt)-OH was coupled directly onto PEG-PS-OH resin (Scheme 19). The Scheme 19

Fmoc and side chain trityl protecting groups were then cleaved under standard conditions. Addition of an aldehyde in acetic acid resulted in imine formation followed by cyclization to provide the support-bound thiazolidine 80. Acylation of the thiazolidine 80 in pyridine with either acetic anhydride, benzoyl chloride, or an isocyanate followed by cleavage with NaOH provided the racemic N-acylthiazolidine 82 as a single diastereomer (epimerization occurs upon hydrolysis). The products were isolated in 97% by HPLC analysis (Table 10). The authors also report that N-acylthiazolidines prepared from electron-rich aromatic aldehydes are not stable in 60-100% TFA/CH2Cl2, conditions that are necessary to cleave many of the side-chain protecting groups Table 10. Thiazolidine-4-carboxylic Acid Derivatives (Scheme 19) derivative

Figure 19. Building blocks used to prepare an ∼500member mercaptoacyl proline library.

entry

R1

R2

yield (%)a

1 2 3 4 5 6 7 8 9 10

C6H4-4-OCH3 C6H4-4-NO2 C6H4-4-OCH3 H H CH(CH3)2 CH(CH3)2 CH(CH3)2 CH2CH(CH3)2 2-thienyl

CH3 CH3 NHCH3 CH3 C6H5 CH3 C6H5 NHC6H5 CH3 CH3

90 0 69 61 81 40 79 58 75 63

a

Crude mass balance after extractive workup.

Synthesis and Applications of Small Molecule Libraries

Chemical Reviews, 1996, Vol. 96, No. 1 575

that are commonly used in solid-phase synthesis. To address this problem, oxidation of the thioether to the sulfoxide was accomplished in good yield using MCPBA, providing acid-stable products. A variety of conditions, however, failed to provide the corresponding sulfone in good yield. 4-Thiazolidinones and Related 4-Metathiazanones. Holmes and co-workers at Affymax have reported the synthesis of 4-thiazolidinones and related metathiazanones from a support-bound amino acid, an aromatic aldehyde and a mercaptoacetic acid derivative (Scheme 20).95 After Fmoc deprotection Scheme 20

Figure 21. Preparation of imidazoles.

Imidazoles. Sarshar and co-workers at Ontogen corporation have developed a method for the solidphase synthesis of imidazole derivatives based on a one-pot procedure.96 The imidazole derivatives may be prepared by either a three-component or a fourcomponent process (Figure 21). In the three-component process NH4OAc, a 1,2-dione, and an aldehyde are heated in AcOH to provide the imidazole product, and in the four-component process NH4OAc, a primary amine, a 1,2-dione, and an aldehyde are submitted to the same reaction conditions (the relative stoichiometry of the reagents is used to control primary amine introduction). The authors perform the synthesis on solid support by linking either the amine or aldehyde component to the support (Wang resin) as shown in Scheme 21. Scheme 21

of commercially available amino acid-loaded resins (polystyrene, PEG-PS, and a polydimethylacrylamide/polyhipe support were all used with equal success), a one-pot reaction was performed with an aryl aldehyde and a mercaptoacetic acid (mercaptoacetic acid and thiolactic acid were used) with heating at 70 °C and with 3 Å molecular sieves as a dehydrating agent. Cleavage from the support with 50% TFA in CH2Cl2 provided the thiazolidinone derivatives in good purity as estimated by HPLC (84-98%, av 94%, Table 11). Diastereomeric thiazolidinones are formed from chiral amino acids, and bulky substituents were shown to provide modest diastereoselectivity (e4:1 major/minor). Attempts to use β-mercaptopropionic acid to form the analogous six-membered ring metathiazanones were not as successful. A derivative of glycine was isolated in good purity, but extension to Ala and Phe led to the formation of 95 >95

a

Yields of unpurified material are based on resin-bound aryl halide as the limiting reagent. All derivatives were of >90% purity as determined by 1H NMR and HPLC analysis.

of iodo- or bromo substituted benzoic acids were loaded onto Merrifield resin (chloromethylated polystyrene) by esterification under standard alkylation conditions (Cs2CO3, KI, DMF). Coupling reactions were performed with a variety of palladium catalysts and boronic acids to produce the biphenyl derivatives 167, which were cleaved from the support using catalytic NaOMe in MeOH/THF. Frenette found that the optimal reaction conditions were 5 mol % Pd(PPh3)4 in DME with 2 M Na2CO3 as a base. Crude mass balances for the reaction sequence employing a variety of substituted aryl boronic acids and halobenzoic acids were generally greater than 95%, and the compounds were obtained in greater than 90% purity as estimated by HPLC analysis (Table 16). Stille Reaction. Deshpande has examined the Stille coupling reaction on support in the synthesis of 4-substituted benzamides.133 4-Iodobenzoic acid was coupled onto Rink resin or Ala-derivatized Wang resin to provide the support-bound aryl iodide (Scheme 37). Stille reaction using a number of stannanes was Scheme 37

then performed with 5 mol % Pd2(dba)3 and added Ph3As to afford the support-bound styrene and biaryl products. Cleavage from support using 5% TFA in CH2Cl2 (Rink) or 90% TFA in CH2Cl2 (Wang) provided the products 170 or 171 in crude mass balances of 85-92% and in greater than 90% purity as estimated by HPLC (Table 17). Table 17. Stille Reaction Products 170 and 171 (Scheme 37) entry

producta

vinyl stannane

yield (%)b

1 2 3 4 5 6 7

170 170 170 170 170 171 171

CHdCH2 (Z)-CHdCH(CH3) CHdC(CH3)2 (E)-CHdCH(Ph) (E)-CHdCH(Ph-3,4-di-OCH3) CHdCH2 (Z)-CHdCH(CH3)

89 91 85 89 90 92 88

a Compounds 170 were synthesized on Rink resin and were isolated as the primary amide. Compounds 171 were synthesized on Wang-Ala resin. b Yields are based on resin-bound 4-iodobenzoic acid as the limiting reagent.

Table 18. Compounds Synthesized by a Heck Reaction (Scheme 38) entry

reactant

product

mass balance (%)a

1 2 3 4 5 6 7 8

(4-carbomethoxy)styrene phenylacetylene ethyl acrylate ethyl propenoate Ph-I 3-bromonaphthyl 2-bromothienyl 3-bromopyridyl

175 175 175 see b below 173 173 173 173

90 90 91 see b below 81 64 76 87

a Mass balances are based on resin-bound 4-vinyl- or 4-iodobenzoic acid as the limiting reagent. The products are >90% pure as determined by 1H NMR and HPLC analysis. b A mixture of products was obtained.

Heck Reaction. Yu and co-workers have examined the generality of the Heck reaction on both support-bound iodides and support-bound alkenes under a variety of reaction conditions (Scheme 38).134 Yields of the final products 173 were excellent when the support-bound aryl iodide was coupled with different alkene components in solution using Pd(OAc)2 in DMF at 80-90 °C. The polymer-bound alkene was also coupled with four different aryl halides using Pd2(dba)3 and P(2-tolyl)3 in DMF at 100 °C to give good yields of the final product 175, although Heck reactions with aryl triflates were not successful (Table 18). Zhou and co-workers have extended the scope of the Heck reaction on solid support to include reactions that proceed under mild conditions (Scheme 39).135 Using a phase-transfer system developed by Scheme 39

Jeffery136 and PEG-PS resin from Millipore, the authors have shown that Heck reactions can be performed on support-bound 4-iodobenzoic acid in aqueous solvent comixtures (DMF/H2O/Et3N, 1:1:1) using Pd(OAc)2, PPh3, and Bu4NCl at 37 °C. Yields of Heck products range from fair to excellent (5495%, av 79%, 6 compounds) for a variety of vinylic reagents. 4. Amide Bond Formation on a Cyclic Template. Lebl and co-workers have synthesized an all-ciscyclopentane template for the display of amide-based functionality.137 Compound 179 is synthesized in solution by a multistep route starting from com-

Synthesis and Applications of Small Molecule Libraries Scheme 40

Chemical Reviews, 1996, Vol. 96, No. 1 587 Table 19. Biaryl Derivatives 184 or 186 (Scheme 41) entry

alcohol or phenol R1

product

yield (%)a

purity (%)b

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

H 4-CH3 4-OCH3 4-OPh 4-Br 4-CO2CH3 4-CN 2-CH3 2-Ph 2-CH3, 4-CHO C6H4-4-Br C6H4-4-CO2CH3 see c below Bu (3-C6H5)-C3H6

184 184 184 184 184 184 184 184 184 184 186 186 186 186 186

75 79 75 81 90 92 99 72 99 80 87 77 94 66 68

92 90 92 89 93 96 95 90 97 94 92 97 98 88 81

a Yields of unpurified material are based on resin-bound aryl halide as the limiting reagent. b Purities were estimated by HPLC analysis. c 4-[(2-aminomethyl)phenyl]phenyl.

mercially available cis-5-norbornene-endo-2,3-dicarboxylic anhydride (Scheme 40). The anhydride is opened with a resin-bound secondary amine of SCAL138-derivatized PEG-PS resin 178 to give a 9:1 mixture of the regioisomeric carboxylic acids (only the major diastereomer is shown). After coupling the resulting free carboxylic acid to an amine using (benzotriazol-1-yloxy)tris(dimethylamino)phosphonium hexafluorophosphate (BOP), the methyl ester is saponified with 0.5% NaOH. The resulting carboxylic acid is then coupled with either a primary or secondary amine and BOP. The N-Boc group is then removed by treatment with TFA, and the resulting amine is acetylated. Final cleavage from the SCAL linker is accomplished by treatment with 1 M TMSBr and 1 M thioanisole in TFA. The authors report that several derivatives with different side chains were obtained in high yield and purity employing the described sequence, but no structures or analytical data were provided. The Selectide group has recently extended this chemistry to include a related scaffold, 1,3,5-trimethyl-1,3,5-cyclohexanetricarboxylic acid (Kemp’s triacid).139 5. Ether Formation. Rano and Chapman have optimized the Mitsunobu coupling reaction on solid support for the synthesis of aryl ethers with either the phenol or alcohol serving as the support-bound component (Scheme 41).140 Either 4-(hydroxymethyl)benzoic acid or 3-(4-hydroxyphenyl)propionic acid was coupled to Rink-derivatized PEG-PS resin with EDC to give the support-bound alcohol or phenol. Scheme 41

Mitsunobu reaction with 5 equiv of both N,N,N′,N′tetramethylazodicarboxamide (TMAD) and Bu3P and excess of the phenol or alcohol followed by cleavage with aqueous TFA provided Mitsunobu products 184 or 186. Crude mass balances ranged from 66-99% (av 82%, 15 compounds) and purities ranged from 81-99% (av 92%) as estimated by HPLC (Table 19). Mitsunobu reactions using diethyl or diisopropyl azodicarboxylate were also successful, but resulted in slightly lower purities and yields of the final products. Krchnak and co-workers have also reported a method for the modification of a resin-bound phenol using the Mitsunobu reaction (Scheme 42).141 OScheme 42

Alkylation of Ac-Tyr-OH and 4-(hydroxybenzoyl)glycine bound to PEG-PS resin was accomplished using PPh3 and DEAD with a variety of primary and secondary alcohols. The Mitsunobu reaction is exothermic; however, reaction mixture warming promotes decomposition of DEAD, liberating ethyl alcohol and producing the corresponding ethyl ether as a byproduct. Accordingly, the reaction is performed by premixing the resin, alcohol, and PPh3, and then adding a solution of DEAD in THF portionwise over the period of 20 min. DIAD was also found to give better results in problem cases, however, the reaction time is longer, typically 3 h instead of 1 h. The product purities by HPLC range from fair to excellent, although yields were not reported (Table 20). This chemistry has been used to synthesize a library of 4200 compounds using split synthesis. Twenty amino acids were coupled to PEG-PS resin, a variety of 10 aromatic hydroxy acids were coupled under standard conditions, and a Mitsunobu reaction was

588 Chemical Reviews, 1996, Vol. 96, No. 1

Thompson and Ellman

Table 20. Alkyl Aryl Ethers 188 or 190 (Scheme 42)

a

product purity (%)a,b

ethyl ether (%)a,b

entry

alcohol R1

188

190

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

methanol ethanol 2-propanol 1-butanol allyl alcohol 1,3-propanediol benzyl alcohol 4-methoxybenzyl alcohol 4-(methylthio)benzyl alcohol 2-(hydroxymethyl)furan 3-(hydroxymethyl)furan 2-(hydroxymethyl)thiophene 4-methyl-5-(2-hydroxyethyl)thiazole 2-(hydroxymethyl)pyridine 3-(hydroxymethyl)pyridine 4-(hydroxymethyl)pyridine 2,6-bis(hydroxymethyl)pyridine 1-(2-hydroxyethyl)pyrrolidine 1-(2-hydroxyethyl)-2-pyrrolidinone 1-(Boc-amino)ethyl alcohol 3-(Fmoc-amino)propyl alcohol

95 96 93 75 95 94 99 98 78 66 98 94 59 84 86 57 82 43 (60) 42 39 (52) 90

99 99 99 80 99 99 93 99 90 74 99 93 85 (96) 99 99 92 98 53 (82) 47 (85) 51 (82) 69

188

189